In December 2003, Fred Reichheld published an article in the Harvard Business Review titled "The One Number You Need to Grow." The article made a remarkable claim: after years of research across hundreds of companies, Reichheld had identified a single survey question that correlated more strongly with business growth than any other customer feedback metric. That question was: "How likely is it that you would recommend our company to a friend or colleague?"
The resulting metric — the Net Promoter Score — became one of the most widely adopted management tools in business history. Within a decade, two-thirds of Fortune 1000 companies were measuring it. Bain and Company built a consulting practice around it. Business books praised it as the holy grail of customer loyalty measurement. Executives received bonuses tied to it.
It also generated some of the most pointed methodological criticism in the history of applied market research. Understanding both what NPS measures and what it does not is essential for anyone who encounters it at work — which is, increasingly, almost everyone.
How Net Promoter Score Works
The mechanics of NPS are straightforward. A customer is asked one question:
"On a scale of 0 to 10, how likely is it that you would recommend [Company/Product] to a friend or colleague?"
Responses are categorized into three groups:
| Score | Category | Description |
|---|---|---|
| 9-10 | Promoters | Enthusiastic customers likely to fuel growth through referrals and repeat purchases |
| 7-8 | Passives | Satisfied but unenthusiastic customers who are vulnerable to competitive offerings |
| 0-6 | Detractors | Unhappy customers who may actively discourage others from using the product |
The NPS is calculated by subtracting the percentage of Detractors from the percentage of Promoters:
NPS = % Promoters - % Detractors
Passives are excluded from the calculation entirely. The resulting score ranges from -100 (all respondents are Detractors) to +100 (all respondents are Promoters).
Reichheld published this framework in a 2003 HBR article and expanded it in the 2006 book The Ultimate Question: Driving Good Profits and True Growth, co-authored with Rob Markey. Bain and Company, where Reichheld is a fellow, commercialized the NPS system, creating certification programs, industry benchmarks, and consulting services built around its implementation.
What NPS Attempts to Measure
The theoretical foundation of NPS rests on a distinction Reichheld drew between two types of profit: bad profits, earned by exploiting or trapping customers in ways that damage loyalty, and good profits, earned by delivering genuine value that customers want to pay for and recommend.
Reichheld argued that customer recommendation behavior — captured by the likelihood-to-recommend question — is the behavioral proxy that best distinguishes companies that earn good profits from those that earn bad ones. His 2003 analysis of customer loyalty data from multiple industries claimed to find that the likelihood-to-recommend question outperformed longer satisfaction surveys in predicting business growth.
"The percentage of customers enthusiastic enough to refer a friend or colleague — perhaps the ultimate act of loyalty — turns out to be the best predictor of growth." — Fred Reichheld, Harvard Business Review, December 2003
The appeal of NPS beyond its predictive claims was its simplicity. A single question requires less customer effort to answer, produces higher response rates, and is easier to benchmark against industry peers than a 30-question satisfaction survey. In an era when organizations were drowning in customer feedback data they lacked the capacity to act on, the promise of one number was genuinely attractive.
The Criticisms: What the Research Actually Found
The most rigorous academic challenge to Reichheld's claims came in a 2007 meta-analysis by Timothy Keiningham and colleagues published in the Journal of Marketing. The study analyzed data from 21 longitudinal studies covering multiple industries and countries and asked a simple question: does NPS predict business growth better than other loyalty metrics?
The answer was no.
Keiningham et al. found that customer satisfaction and other loyalty metrics predicted revenue growth as well as or better than NPS in most of their analyses, and that NPS did not hold a statistically consistent advantage across industries or contexts. The specific claim that likelihood to recommend is the single best predictor of growth was not supported by the data.
This finding has been replicated and extended in subsequent research. A 2020 paper by Daniel Markovitz and colleagues examined customer data from subscription businesses and found that actual renewal behavior — not NPS — was the strongest predictor of retention.
Specific Methodological Criticisms
Beyond the predictive validity question, NPS has attracted a range of methodological criticisms from survey researchers:
The 11-point scale has questionable interval properties. The difference between a 6 and a 7 (which crosses the Detractor/Passive boundary) is treated as qualitatively different from the difference between a 5 and a 6, even though respondents presumably treat adjacent numbers as roughly equivalent. The categorization discards information from the raw scale.
Response set effects vary by culture. Research on cross-cultural survey research documents that respondents in different countries use rating scales differently: American respondents tend toward higher scores, European respondents toward more moderate ones. This makes international NPS comparisons unreliable without cultural calibration.
The metric is highly sensitive to survey methodology. NPS measured by phone differs from NPS measured by email, which differs from NPS measured by in-app pop-up. The timing of the survey relative to the customer experience has large effects. These context effects make it difficult to compare scores across channels or over time if the methodology has changed.
Tying compensation to NPS encourages gaming. When bonuses, performance reviews, or team rankings depend on NPS, organizations reliably see scores improve — through survey cherry-picking, coaching customers before the survey, or selecting the survey sample to exclude likely detractors. The metric that was supposed to reveal the truth about customer experience becomes a managed number rather than a genuine signal.
What NPS Does and Does Not Predict
Understanding NPS requires distinguishing between what the metric has been shown to correlate with and what it has not.
| NPS May Predict | NPS May Not Predict |
|---|---|
| Directional trends in customer loyalty over time | Specific revenue growth better than alternatives |
| Likelihood of customer churn when scores are very low | Customer lifetime value at an individual level |
| General relationship health between company and customer | Future purchase behavior with consistent reliability |
| Employee and customer perception of brand | Actual word-of-mouth referral behavior |
The last row is particularly significant. Reichheld's original hypothesis was about actual recommendation behavior — customers physically recommending the product to others. But NPS measures stated likelihood to recommend, not actual recommendation behavior. Research by Morgan and Rego (2006) found that actual referral behavior is a poor proxy for NPS scores, and that satisfied customers who give high NPS scores do not necessarily generate more referrals.
Alternatives and Complementary Metrics
The criticisms of NPS have produced a small industry of alternative metrics. The most widely adopted are:
Customer Satisfaction Score (CSAT)
CSAT measures satisfaction with a specific interaction or experience, typically using a 1-5 or 1-10 scale immediately after the interaction. It answers the question: "How satisfied were you with this interaction?"
Strengths: More specific than NPS; directly tied to particular touchpoints; widely understood.
Weaknesses: Measures satisfaction in the moment, not relationship quality; satisfaction and loyalty are not the same thing; satisfaction with individual transactions can be high while overall loyalty is low.
Customer Effort Score (CES)
Customer Effort Score was developed by the Corporate Executive Board (now Gartner) and published in a 2010 Harvard Business Review article titled "Stop Trying to Delight Your Customers." CES measures how much effort a customer had to exert to resolve an issue or complete a task, typically using a scale from "very low effort" to "very high effort."
The CEB's research found that reducing customer effort predicted customer retention better than increasing customer delight in service contexts — a counterintuitive finding that challenged the prevailing emphasis on exceeding expectations. Making it easy to cancel, return products, or resolve billing issues reduced churn more effectively than trying to provide exceptional experiences on the way in.
Strengths: Strong predictor of retention in service contexts; actionable (the question of "where did you have to work hard?" directly identifies process failures).
Weaknesses: Less useful for relationship measurement; primarily applicable to service and support interactions.
Combining Multiple Metrics
Most measurement experts now recommend using multiple customer metrics rather than any single number:
- NPS for relationship-level sentiment and competitive benchmarking
- CSAT for transaction-level quality feedback
- CES for identifying friction in service and support
- Churn rate and repeat purchase rate as behavioral measures that do not depend on survey response
The combination approach acknowledges that no single metric captures the full complexity of customer experience, while avoiding the paralysis of tracking too many numbers simultaneously.
NPS Industry Benchmarks
NPS scores vary dramatically by industry, making absolute score comparisons across sectors misleading. A score of 30 might be excellent in one industry and mediocre in another.
| Industry | Typical NPS Range | Notes |
|---|---|---|
| Software / SaaS | 30-60 | Wide variance by product type |
| Consumer electronics | 30-50 | Apple, Samsung drive high end |
| Financial services (banking) | 20-40 | Highly regulated, competitive |
| Airlines | 0-40 | High variance; loyalty programs inflate scores |
| Internet / cable providers | -10-20 | Structural low satisfaction |
| Healthcare | 20-40 | Complex; varies by specialty |
| Retail | 30-50 | E-commerce typically higher than in-store |
Bain and Company publishes annual industry benchmark reports that provide more granular segmentation. Comparing your NPS to industry peers is more meaningful than comparing it to an absolute scale, but even industry comparisons require controlling for survey methodology, customer segment, and geographic market.
Best Practices for Using NPS Responsibly
Given the criticisms and the evidence, how should organizations use NPS?
Use it as a signal, not a verdict. NPS is most valuable as a directional indicator that prompts investigation, not as a precise performance measurement. A declining NPS score tells you that something has changed in the customer relationship; it does not tell you what or what to do about it.
Always follow up with the "why." The single-question format is NPS's practical advantage but its analytical weakness. Without open-ended follow-up questions asking respondents to explain their score, you know your NPS but not the reasons behind it. The follow-up question ("What is the primary reason for your score?") is where the actionable insight lives.
Segment your results. An overall NPS of 35 may conceal a score of 55 among enterprise customers and 15 among SMB customers — two completely different businesses in the same number. Segment by customer type, product line, channel, tenure, and geography to find where the relationship is strong and where it is breaking down.
Track trends, not absolutes. The comparison that matters most is your score this quarter versus last quarter, or this year versus last year. Absolute NPS values are too sensitive to methodology and cultural context to be reliable across companies; trends within a consistent methodology are more informative.
Never tie individual employee compensation directly to NPS. The research on this is unambiguous: financial incentives tied to satisfaction scores reliably inflate the scores without improving the underlying experience. Use NPS at the team or division level, and pair it with behavioral metrics that are harder to game.
Close the loop with detractors. One of the highest-return investments an NPS program can make is direct outreach to customers who gave scores of 0-6. Understanding their specific experience, acknowledging the problem, and attempting to resolve it has been shown to increase retention rates among at-risk customers and provides a direct source of improvement priorities.
Net Promoter Score is not the one number you need to grow. It is one useful signal among several, limited by the same survey constraints as any other metric, and best used as a prompt for deeper investigation rather than a destination in itself. Used with those limitations in mind, it can be a valuable tool. Used as a substitute for genuine customer understanding, it becomes another number that people optimize rather than an experience that people improve.
Frequently Asked Questions
What is Net Promoter Score (NPS)?
Net Promoter Score is a customer loyalty metric based on a single question: 'How likely are you to recommend this company to a friend or colleague?' Respondents answer on a 0-10 scale. Those scoring 9-10 are Promoters, 7-8 are Passives, and 0-6 are Detractors. NPS is calculated by subtracting the percentage of Detractors from the percentage of Promoters, producing a score from -100 to +100. Fred Reichheld introduced the concept in a 2003 Harvard Business Review article titled 'The One Number You Need to Grow.'
What is a good NPS score?
NPS benchmarks vary significantly by industry, making cross-company comparisons meaningful only within sectors. Generally, any positive NPS is considered acceptable, above 50 is considered excellent, and above 70 is world-class. Consumer technology companies and subscription services tend to post higher scores than utilities or healthcare. Bain and Company, which commercialized NPS, publishes annual industry benchmarks that show median scores ranging from below 20 in some industries to above 60 in others.
What are the main criticisms of NPS?
The most rigorous critique comes from a 2007 meta-analysis by Timothy Keiningham and colleagues published in the Journal of Marketing, which analyzed 21 longitudinal studies and found that NPS was not consistently superior to other satisfaction and loyalty metrics in predicting revenue growth — contradicting Reichheld's central claim. Additional criticisms include: the 11-point scale has weak interval properties; the Promoter/Passive/Detractor categorization discards information; scores are highly sensitive to survey method, timing, and market context; and the metric encourages gaming.
What are alternatives to NPS?
Customer Satisfaction Score (CSAT) measures satisfaction with a specific interaction using a simple rating scale and is better suited for transactional feedback. Customer Effort Score (CES), developed by the Corporate Executive Board in 2010, measures how much effort a customer had to exert to resolve an issue and has shown strong correlation with customer retention in service contexts. Some researchers advocate for combining multiple metrics — NPS for relationship tracking, CSAT for transaction quality, and CES for service efficiency — rather than relying on any single number.
How should companies use NPS responsibly?
NPS is most useful as a directional signal and conversation starter rather than a precise performance metric. Best practices include following up with open-ended questions to understand the 'why' behind scores, segmenting results by customer type and journey stage, tracking trends over time rather than absolute values, closing the loop with detractors through direct outreach, and never using NPS as the sole input for major strategic decisions. Tying employee compensation to NPS scores tends to produce gaming and score inflation rather than genuine loyalty improvement.